4 ms·

I wonder why their old system couldn't scale more horizontally so that there's <50 req/s per service instance in order for the system to GC properly Although t

I wonder why their old system couldn't scale more horizontally so that there's <50 req/s per service instance in order for the system to GC properly Although to be fair <50 req/s/instance is pretty bad performance so a rewrite is probably warranted

ec109685 5y ago
You can get unlucky and have multiple slow page requests land on the same NodeJS worker and interfere with each other, so as a result have to massively over scale to reduce chances. With their new model, they distribute requests explicitly to workers so they can’t interfere with each other and drop server side rendering if they don’t have capacity.